View
7
Download
0
Category
Preview:
Citation preview
Moral Hazard and Peer Monitoring in a Laboratory Microfinance Experiment*
Timothy N. Cason†, Lata Gangadharan‡ and Pushkar Maitra§
February 2007
Abstract
Most problems with formal sector credit lending to the poor in developing countries can be attributed to the lack of information and inadequate collateral. One common feature of successful credit mechanisms is group-lending, where the loan is advanced to an individual if he/she is a part of a group and members of the borrowing group can monitor each other. Since group members have better information about each other compared to lenders, peer monitoring is less expensive than lender monitoring. Theoretically this leads to greater monitoring and greater rates of loan repayments. This paper reports the results from a laboratory experiment of group lending in the presence of moral hazard and (costly) peer monitoring. We compare peer monitoring treatments when credit is provided to members of the group sequentially and simultaneously, and a case of individual lending with bank monitoring. Our results suggest that peer monitoring results in higher loan frequencies, higher monitoring and higher repayment rates compared to bank monitoring. Although the dynamic incentives provided by sequential leading generate the greatest equilibrium efficiency, simultaneous group leading provides equivalent empirical performance. JEL Classification: G21, C92, O2. Key words: Group Lending, Monitoring, Moral Hazard, Laboratory Experiment, Loans, Development
* We have benefited from extensive discussions with Dyuti Banerjee and from comments by Chikako Yamauchi, participants at the Experimental Economics Workshop at the University of Melbourne, Behavioral Economics working group at Monash University, ESA International Meetings in Atlanta, ESA European Meetings in Nottingham, the NEUDC Conference at Cornell University and seminar participants at Jadavpur University, Kolkata and at the Australian National University, Canberra. We would like to thank Gautam Gupta for his assistance in organizing the sessions at Jadavpur University, Kolkata. Tania Dey, Simon Hone, Vinod Mishra and Roman Sheremeta provided excellent research assistance. The usual caveat applies. † Timothy Cason, Department of Economics, Krannert School of Management, Purdue University, West Lafayette, IN 47907-2076, USA. Email: cason@purdue.edu ‡ Lata Gangadharan, Department of Economics, University of Melbourne, VIC 3010, Australia. Email: latag@unimelb.edu.au § Pushkar Maitra, Department of Economics, Monash University, Clayton Campus, VIC 3800, Australia. Email: Pushkar.Maitra@Buseco.monash.edu.au
1. Introduction
There now exists a significant body of research that examines the failure of formal sector
credit lending programs aimed at the poor in developing countries. Evidence of this failure is
shown in the inability to reach target groups and low overall repayment rates. This failure is
attributed primarily to asymmetric information (both adverse selection and moral hazard) and
inadequate enforcement.1
The last few decades have, however, witnessed the development of innovative and
highly successful mechanisms for the provision of credit to the poor. The most common of
these is group-lending. Rather than the bank (or the lender) making a loan to an individual
who is solely responsible for its repayment, the bank makes a loan to an individual who is a
member of a group and the group is jointly liable for each member’s loans. In particular, if
the group as a whole is unable to repay the loan because some members default on their
repayment, all members of the group are ineligible for future credit. The Grameen Bank in
Bangladesh is possibly the most well known of such group lending programs. The repayment
rate in this lending scheme is around 92 percent, and less than 5 percent of loan recipients
were outside the target group (Morduch, 1999). The success of the Grameen Bank has led
policy makers and NGO’s around the world to introduce similar schemes. Around 100
million people are estimated to have participated in some form of a microfinance project (see
Gine, Jakiela, Karlan and Morduch, 2005). The 2006 Nobel Prize for Peace to microfinance
pioneer Muhammed Yunus has also put the success of microfinance in the world spotlight.
Micro-lending is also moving from non-profit towards a profit-making enterprise, with big
banks such as Citigroup now backing such loans (Bellman, 2006).2
1 For example, it has been argued that the percentage of ineligible beneficiaries in the Integrated Rural Development Program (IRDP) in India, one of the largest programs of provision of formal sector credit to the poor in the world, was between 15 and 26 percent, with the highest reported being 50 percent. The repayment rate for IRDP loans was only about 40 percent for the whole country (see Pulley, 1989). 2 While microfinance programs are most widespread in less developed countries they are by no means confined to them. Microfinance programs have been introduced in transition economies like Bosnia and Russia and even
2
The success of these group lending programs arises, in part, because they are better
able to address the enforcement and informational problems that generally plague formal
sector credit in developing countries. Group lending programs help solve the enforcement
problem through peer monitoring. Stiglitz (1990) and Varian (1990) argue that since group
members are likely to have better information compared to an outsider (the bank), peer
monitoring is relatively cheaper compared to bank monitoring, leading to greater monitoring
and hence greater repayment. Banerjee, Besley and Guinnane (1994) argue that explanations
based on peer monitoring do a better job of explaining the success of group lending programs
than other explanations. Ghatak and Guinnane (1999) develop a model of moral hazard and
monitoring and find that if the social sanctions are effective enough or if monitoring costs are
low enough, the joint liability provided by group lending improves repayment rates.
Chowdhury (2005), by contrast, is less optimistic. He finds that in the absence of sequential
financing or lender monitoring, group lending programs will typically involve under-
monitoring with the borrowers investing in undesirable projects.3
The empirical evidence on these issues, unfortunately, is rather limited. The
theoretical propositions and results are often supported by anecdotal evidence but these
results have not been established as empirical regularities. In recent years researchers have
called for well designed economic experiments to help examine the roles of various
mechanisms that drive performance in microfinance programs (Morduch, 1999, Armendariz
de Aghion and Morduch, 2005).
The aim of this paper is to understand specific aspects of group lending schemes,
using controlled experimental methods. We report the results from a laboratory experiment of
in developed countries like Australia, Canada and the US. See for example Conlin (1999), Armendariz de Aghion and Morduch (2000, 2005), and Fry, Mihajilo, Russell and Brooks (2006). 3 How group lending solves the problem of adverse selection is analysed by Ghatak (1999, 2000), Van Tassel (1999), Armendariz de Aghion and Gollier (2000). The argument is based on endogenous group formation (and positive assortative matching): group lending with joint liability will result in self selection with safe borrowers clubbing together and screening out risky borrowers.
3
group lending in the presence of moral hazard and (costly) peer monitoring.4 We compare
treatments when credit is provided to members of the group sequentially and simultaneously,
as well as a benchmark treatment when loans are given to individuals and monitored by
lenders.
Our work complements the growing body of research that can broadly be
characterized as field experiments in microfinance (see for example Gine, Jakiela, Karlan and
Morduch, 2006; Gine and Karlan, 2006, Kono, 2006, Cassar, Crowley and Wydick, 2007).
The laboratory approach that we use in this paper can address issues in different ways
compared to field experiments. It is difficult to vary specific properties of institutions in
controlled experiments in the field due to problems of replicability, data accessibility and
comparability (see for example Bolnik, 1988 and Hulme, 2000). Furthermore some relevant
variables, such as actual monitoring costs, remain unobserved. The laboratory approach on
the other hand can help us control for specific parameters and observe behavior in simulated
microfinance institutions. In our case it can help in isolating and clarifying the impact of
different design features on repayment rates and project choice, by implementing an
environment that is carefully aligned with the theoretical models relating to moral hazard and
peer monitoring in microfinance programs. Of course, the laboratory approach has some
drawbacks. For example, while the laboratory experiment included human subject behavior,
the subjects are university students making decisions for relatively low stakes.5 In field
experiments, by contrast, participants are often the actual borrowers who are involved in
microfinance programs. This advantage of field experiments comes at the cost of some loss
of experimental control, however. For example, spillover effects could exist from one village
to another or from the treatment group to the control group, creating more noise in the data. 4 In this paper, we focus on informational asymmetries due to moral hazard and not due to adverse selection. In particular we restrict ourselves to exogenously formed groups (with random re-matching) and leave the issue of endogenous group formation (positive assortative matching) for future research. 5 We do, however, employ subjects both from a developed (Australia) and a developing (India) country to measure possible subject pool effects, and find virtually none.
4
Laboratory experiments that examine the impact of specific design features on
performance of microfinance models are rare. Abbink, Irlenbusch and Renner (2006) and
Seddiki and Ayedi (2005) both examine the role of group selection in the context of group
lending. Both experiments are designed as investment games where each group member
invests in an individual risky project whose outcome is known only to the individual, and
both find that self-selected groups have a greater willingness to contribute. Neither of these
papers analyze the role of peer monitoring.
Our experiment examines several aspects of group lending programs. The first is the
argument that sequential lending is crucial to the success of group lending schemes. The
Grameen Bank, for example, adopts this kind of a lending policy: groups have five members
each and loans are initially given to two randomly chosen members, to be repaid in regular
installments over a period of one year. If they pay their initial installments, then two more
borrowers in the group receive the loan and so on. Ray (1998) argues that this kind of
sequential lending minimizes the contagion effect associated with individual default.
Sequential lending can also minimize the potential of coordination failure. Aniket (2006) and
Chowdhury (2005) argue that in a simultaneous group lending scheme with joint liability and
costly monitoring, peer monitoring by borrowers alone is insufficient and that sequential
lending that incorporates dynamic incentives is required.6 Our experiment examines the
empirical validity of these predictions by comparing the performance of sequential lending
and simultaneous lending in the presence of moral hazard and costly peer monitoring
The second issue is whether peer monitoring indeed does better than active lender
monitoring. The bank or the lender in general is an outsider who often has less information
about the borrowers. Therefore monitoring individual borrowers could be very expensive for
6 Dynamic incentives mean that banks make future loan accessibility contingent on full repayment of the current loan to prevent strategic default.
5
the lender. Specifically, we study whether in the presence of moral hazard, group lending
with peer monitoring does better than individual lending with bank monitoring.7
The third issue is the relative benefits of individual and group lending. Over the years
there has been a discernible shift from group lending to individual lending in microfinance
programs and there are a number of theoretical reasons that have been advanced to explain
this shift.8 First, clients often dislike tensions caused by group lending. Second, low quality
clients can free-ride off high quality clients leading to an increase in default rates. Third,
group lending can be more costly for the clients as they often end up repaying the loans of
their peers. Theoretically the results are mixed.9
Our laboratory experiment is able to address each of these issues. Our results show
that group lending (with peer monitoring) performs better compared to individual lending
(with active lender monitoring), reflected in higher loan frequencies and repayment rates.
This occurs even though repayment rates with individual lending considerably exceed the
theoretical prediction, which may reflect social preferences such as reciprocity. These results
differ from those observed in the field by Gine and Karlan (2006) and Kono (2006), who find
high performance in individual lending schemes. Their explanation is based on the argument
of Grief (1994), who argues that a more individualistic society requires less information
among players and is thus able to grow faster.10 We also find that within group lending it
7 Peer monitoring and peer enforcement have been observed to deter free riding in several experiments relating to other social dilemma situations, such as common pool resource environments and the voluntary provision of public goods. See Fehr and Gaechter (2000), Barr (2001), Masclet, Noussair, Tucker and Villeval (2003), Walker and Halloran (2004), and Carpenter, Bowles and Gintis (2006) for experimental evidence. 8 The terms individual and group lending as defined in this paper essentially correspond to the terms individual and group (joint) liability. We use the term group lending to describe the situation where individuals are both borrowers and simultaneously guarantors of their partners’ loans. 9 Armendariz de Aghion and Morduch (2000, 2005) argue that group lending (joint liability) is just one element in successful microfinance schemes. Chowdhury (2005) argues that mere joint liability does not work and he emphasizes the role of dynamic incentives: in his model a combination of joint liability and dynamic incentives work best in terms of project choice and repayment. Che (2002) argues that joint liability schemes create problems of free-riding and worsen repayment rates, but when projects are repeated multiple times, group lending dominates individual lending. Rai and Sjostrom (2004) emphasize the importance of cross-reporting in achieving efficiency in group lending. 10 We would like to point out a specific design feature of the field experiment conducted by Gine and Karlan (2006), which could at least partly explain the different results. In that experiment, the existing field centres with
6
does not matter whether loans are made simultaneously or sequentially. Although the
dynamic incentives provided by sequential lending can improve efficiency relative to
simultaneous group lending, performance is equivalently high in the two group lending
treatments because agents tend to play the efficient equilibrium in the simultaneous case.
2. Theoretical Framework
Consider a scenario where two borrowers require one unit of capital (say $1) each for
investing in a particular project. The bank, which provides this capital in the form of a loan,
can either make the loan to an individual (individual lending) or it can loan to the borrowers
as a group (group lending). Joint liability for the repayment of the loan exists in the case of
group lending. Borrowers can invest in two different types of projects: one project has a large
verifiable income and no non-verifiable private benefit, while the other has a large non-
verifiable private benefit and no verifiable income. The bank prefers the first project, where it
can recoup its investment, but the borrowers prefer the second one. In the absence of
monitoring, the borrowers will choose to invest in the second project and the bank, knowing
this, will choose not to make the loan.
Let us elaborate on the model, which follows Chowdhury (2005) and Ghatak and
Guinnane (1999). Suppose that there are two borrowers: 1B and 2B . Two projects are
available to each borrower: project S (verifiable) and project R (non-verifiable). If Project S is
chosen, the return is H (verifiable by monitoring) and if project R is chosen, then the return is
b (not verifiable) with b H< . The 1 dollar cost of each project is financed by a loan from the
bank (or a lender) since the borrowers do not have any funds of their own. When the two
borrowers ( 1B and 2B ) borrow together as a group, each borrower receives 1 dollar from the group liability loans were converted to individual liability loans. Lenders therefore had prior information about the borrowers’ characteristics from the group lending field sessions and this could be used in the individual lending sessions at no extra cost. As a result the monitoring costs did not necessarily change as they moved from group lending to individual lending. Furthermore, participants had some experience with group lending before branching off on their own in the individual lending schemes. These reasons could have contributed towards the higher performance outcomes in the individual liability programs in their study.
7
lender. The amount to be repaid is ( )1r > in the case of individual lending or 2r in the case
of group lending. We assume that this r is fixed exogenously.
In the case of the individual lending, if the borrower chooses project S the return to
the bank is r ; otherwise it is 0. The return to the borrower is H r− if the borrower chooses
project S, and is b if the borrower chooses project R. We assume that H r b− < so that
borrowers prefer project R. Banks on the other hand prefer project S. In the case of group
lending, if both borrowers choose project S, the return to each borrower is H r− and the
return to the bank is 2r . If both borrowers choose project R, the return to each borrower is b
and the return to the bank is 0. Finally if one borrower chooses project R and the other
chooses project S, then due to joint liability the return to the borrower choosing project S is 0
while that of the borrower choosing project R is b and the return to the bank is H . We
assume that 2H r≤ . In the case of group lending it is therefore in the interest of both the
bank and the borrowers to ensure that the other member of the group chooses project S.
An informational asymmetry arises because each borrower knows the type of his own
project, but the lender or the other borrower in the group (the partner) can find out the
borrower’s project choice only with costly monitoring. The monitoring process works as
follows: Borrower i can, by spending an amount ( )ic m in monitoring costs, obtain
information about the project chosen by the other borrower in his group with probability
im [ ]0,1∈ . This information can be used by borrower i to ensure that the other borrower in
the group chooses project S. The bank (lender) can also acquire this same information by
spending an amount ( )c mλ . We assume that 1λ ≥ in order to capture the notion that peer
monitoring is less expensive than monitoring by the bank. We assume a quadratic monitoring
cost function so that ( )2
2i
imc m = . Monitoring level m costs the bank
2
2mλ . If the
8
borrower i chooses monitoring level im , then with probability im he can force the other
borrower in the group to choose project S. 11
Individual Lending
First consider individual lending (with bank monitoring). There are three stages to the game.
Stage 1: Bank chooses whether or not to lend $1 to the borrower.
Stage 2: Bank chooses the level of monitoring, conditional on deciding to lend.
Stage 3: Borrower chooses either project R or project S.
It is straightforward to solve for the sub game perfect Nash equilibrium of the game
by backward induction. If the bank lends, it chooses m to maximize 2
12mmr λ
− − , which
gives * rm λ= . Therefore the expected return to the bank is 2
12rλ
− , so the bank will provide
the loan if and only if 2
1 02rλ
− > i.e. if 2 2r λ> . This implies that individual lending is
feasible only if the costs of monitoring relative to the return are sufficiently low: 2
2rλ < .
Group Lending: Simultaneous
The sequence of events in group lending is as follows:
Stage 1: Bank chooses whether or not to lend $2 to the group. There is joint liability, so that
if one borrower fails to meet his obligations, then if the other borrower has verifiable income
he must pay back the bank for both borrowers.
Stage 2: The borrowers simultaneously choose the level of peer monitoring, im .
Stage 3: Both borrowers choose either project R or project S.
11 We could think of different ways in which monitoring works in practice: information acquired by the borrowers about each other’s project choice may be passed on to the lender who then uses this information to force the borrowers to choose project S. Alternatively, the borrowers can use some form of social sanctions or peer punishment to ensure that the other borrower chooses project S.
9
Note that here both monitoring and lending is simultaneous and we call this simultaneous
group lending. Again the sub game perfect Nash equilibrium is solved by backward
induction. Borrower i will choose monitoring im to maximize
( ) ( ) ( ) ( )2
1 1 *0 12
ii j j i j j
mm m H r m b m m m b⎡ ⎤ ⎡ ⎤− + − + − + − −⎣ ⎦ ⎣ ⎦
The first order condition is: ( ) 0j im H r m− − = . Likewise the first order condition for
borrower j is: ( ) 0i jm H r m− − = . Clearly * * 0i jm m= = is a Nash equilibrium. We call this
the inefficient (zero-monitoring/zero-lending) equilibrium. In this case there is a strategic
complementarity between the monitoring levels of the two borrowers. A particular borrower
knows that if the other borrower monitors and he does not, then he will end up with a payoff
of 0. If however the other borrower does not monitor then he has no incentive to monitor as
well. Mere joint liability and peer monitoring does not solve the moral hazard problem.
Remember however that [ ]0,1m∈ . Now consider the reaction function
( )i jm m H r= − of borrower i with respect to that of borrower j . Since 1H r− > (the
return on project S exceeds the amount that must be repaid), then there exists a 1j jm m= <
(say) such that the best response is 1im = for j jm m≥ . So the reaction function of borrower
i with respect to that of borrower j can be written as:
( ) )
( for 0,
1 for ,1
j j j
i
j j
m H r m mm
m m
⎧ ⎡− ∈ ⎣⎪= ⎨⎤∈⎪ ⎦⎩
In this case ** ** 1i jm m= = is also a Nash equilibrium. We can call this the efficient
(monitoring/lending) equilibrium. Figure 1 presents the reaction functions for 1.75H r− = . It
is important to note that the reaction functions are upward sloping. We will return to this
issue in our estimation results, below.
10
The bank’s payoffs in these two monitoring game equilibria determine whether it will
lend. For the inefficient (0,0) case, the expected payoff to the bank is 2 0− < and group
lending is not feasible. The payoff to both borrowers in this case is 0 . On the other hand, for
the efficient ** **1 2 1m m= = case, the payoff to the bank is 2 0r > and the payoff to both
borrowers is 12H r− − . Clearly ** **
1 2 1m m= = is the payoff-dominant equilibrium. Although
this also makes it a focal point equilibrium (Schelling, 1980, p. 291), previous experimental
evidence indicates that this is not a sufficient condition for “behavioral” equilibrium selection
(e.g., Van Huyck, Battalio and Beil, 1990).
Group Lending: Sequential
An alternative to simultaneous lending is to lend sequentially to group members with the
order chosen randomly. Here initially only one (randomly chosen) member of the group
receives a loan. Depending on whether this loan is repaid, the bank decides whether or not to
lend to the other member of the group. This incorporates dynamic incentives, which have
become increasingly popular among researchers and practitioners in microfinance. The
sequence of events is as follows:
Stage 1: Bank chooses whether or not to lend $1 to one of the members of the group. It puts
the other dollar into alternative use, which yields r .
Stage 2: The borrowers simultaneously choose their levels of monitoring mi.
Stage 3: One of the borrowers is chosen at random (with probability 0.5) to receive the loan,
if the bank lends. This borrower iB decides whether to invest in R or S.
If iB invests in project R, then he gets b and neither jB nor the bank receives
anything. The game stops here. If iB invests in project S the game continues to round 2. In
this case borrower iB receives H r− and the bank receives r . This amount H r− is
11
invested and gives ( ) 1H r r− < , so that self financing among the borrowers is not possible.
Of course if jB is successful in her monitoring, then iB has to invest in project S.
Stage 4: The game moves to round 2 only if iB (the randomly chosen first borrower) invests
in project S in round 1. The bank lends $1 to jB who invests in either project R or project S
(of course if iB was successful in her monitoring, then jB has to invest in project S).
If jB invests in project R then the bank gets ( )H r r− and jB gets b. If jB invests in
project S, then the bank gets r. Total surplus is ( )( )1H r r− + and this is allocated among the
two borrowers. jB gets ( )( )1H r rα − + and iB gets ( )( )( )1 1H r rα− − + .
There is now positive level of monitoring irrespective of α and group lending (with
joint liability) is a feasible and unique equilibrium outcome. Even though both borrowers still
have an incentive to choose the non-verifiable project R, the sequential lending increases their
incentive to monitor. In this case the reaction functions for the two borrowers are symmetric
and are given by
( )( ){ }( )( ){ }
1 121 12
k l
l k
m b m H r r br
m b m H r r br
α
α
⎡ ⎤= + − + −⎣ ⎦
⎡ ⎤= + − + −⎣ ⎦
Solving out we get
( )( )2 1k l
bm m mb r H r rα
= = =+ − − +
The expected payoff to the bank is
12
k l k lm m r m m rr r
+⎛ ⎞⎛ ⎞− −⎜ ⎟⎜ ⎟⎝ ⎠⎝ ⎠
Therefore a unique and positive level of monitoring exists, irrespective of the value of
α . This positive level of monitoring occurs because even if borrower jB does not monitor,
12
iB has an incentive to monitor. To see this, suppose that jB receives the loan in round 1
(remember that the order of receiving the loan is determined randomly). If iB does not
monitor, jB will invest in project R and then iB will receive a payoff of 0. By choosing a
positive level of monitoring, iB can increase the probability that jB invests in project S in
which case the game continues onto the second round and iB gets the loan. Moreover given
that iB is going to monitor, jB has an even greater incentive to monitor due to the strategic
complementarity of monitoring. So the sequential nature of the lending scheme and the
simultaneous choice of the level of monitoring (before a borrower knows whether he is the
first or the second borrower) leads to an efficient (monitoring/lending) equilibrium, as long as
the equilibrium monitoring levels are sufficient to provide positive returns to the lender.
3. Experimental Design
We conducted 8 sessions in each of the three treatments with 12 subjects in each session. The
288 subjects were graduate and undergraduate students at Monash University and University
of Melbourne, Australia and Jadavpur University, Kolkata, India. We conducted sessions in
two countries to examine whether subjects in India, who are perhaps more exposed to issues
relating to microfinance and who share more cultural similarities to targeted borrowers,
would exhibit behavioral differences from the subjects in Australia.12 All subjects were
inexperienced in that they had not participated in a similar experiment. Compared to the
Australian sample, the Indian sample had a lower proportion of females, a greater proportion
of Business/Economics/Commerce majors, and a higher proportion of subjects that lived in a
major metropolis when they were aged 15. The z-tree software (see Fischbacher, 2007) was
used to conduct the experiment. Each session lasted approximately 2 hours, including
12 Following Muhammad Yunus being awarded the Nobel Peace Prize in 2006, microfinance and Grameen Bank have received considerable media attention in India and in particular in Kolkata, which has cultural and linguistic similarities to Bangladesh.
13
instruction time. Subjects earned AUD 25 – 35 or its purchasing power equivalent on
average.13 The instructions (included for the simultaneous lending treatment 2 in the
appendix) used the borrowing and lending terminology employed in this description.
We employed three treatments to examine the equilibrium predictions described in
Section 2. In treatment 1 the 12 subjects were randomly divided into groups of two with each
group consisting of one borrower and one lender. In treatments 2 and 3 the 12 subjects were
randomly divided into groups of three with each group consisting of two borrowers and one
lender. The role of each subject (as a borrower or as a lender) was determined randomly and
remained the same throughout each session, which ran for 40 periods. At the end of every
period participants were randomly re-matched. After reading the instructions and before the
actual session began, the participants answered a set of questions relating to the instructions
and they were paid in cash (at the end of the experiment in addition to their earnings from the
actual experiment) $0.50 or Rs 5.00 for each correct answer.
The two projects available to borrowers, S and R, each cost $1, to be financed by a
loan from the lender. In the individual lending treatment, the lender chose whether or not to
invest $1 into this loan. In the group lending (simultaneous and sequential) treatments, the
lender chose whether or not to invest $2 into the loan ($1 to each borrower). The lender could
choose to make the loan to both borrowers or to neither. She could not make a loan to only
one borrower in the group. If the lender chose not to make the loan, she earned $1.50 (or
$0.75 in the individual lending treatment) for the period. In the group lending treatments, if
the borrower received the loan, he could monitor the project choice of the other borrower in
the group by choosing to pay a monitoring cost (C). Both borrowers could monitor each
other. If borrower X incurred a cost C on monitoring, there was a chance of M that the other
borrower Y would automatically be required to choose project S. Otherwise the other
13 At the time of the experiment, 4 Australian dollars were worth about 3 U.S. dollars.
14
borrower could choose either project R or project S. In the sequential lending treatment, the
borrowers were randomly determined to be the first or the second borrower in the group to
receive the loan. In this case if the first (randomly chosen) borrower’s actual project choice
was R, then the lender’s second dollar was automatically allocated to her savings account
where she earned $0.75 for this dollar. All monitoring decisions were made simultaneously.
We used the strategy method to elicit decisions from the borrowers.14 The use of this
method implies that the borrowers and lenders made decisions simultaneously and borrowers
made their decision before they knew whether or not they had received the loan. In the case
of sequential lending, the borrowers made monitoring decisions before they knew whether
they were the first or the second borrower in their group to receive a loan. They did, however,
know whether they were the first or the second borrower to receive the loan at the time of
making their project choice. Panel A of Table 1 presents the relationship between C and M
for treatments 2 and 3. Table 2 presents the earnings of the borrowers and the lender under
the different project choice scenarios. Note that the even division for two S projects in the
sequential lending (Panel B) indicates our choice of α = 0.5.
In the individual lending treatment, if the lender decided to invest $1 in a period
(make the loan), she could monitor the project choice of the borrower by choosing to pay a
monitoring cost (C). Panel B of Table 1 presents the relationship between C and M in this
case, which is based on λ = 4.5. Lenders paid their selected monitoring costs whenever they
made the loan, regardless of whether or not the monitoring was successful. If unsuccessful,
the borrower could choose either project S or project R. All decisions were revealed to all
members of the two- or three- person group at the end of each period.
4. Hypotheses to be Tested 14 The strategy method simultaneously asks all players for strategies (decisions at every information set) rather than observing each player’s choices only at those information sets that arise in the course of a play of a game. This allows us to observe subjects’ entire strategies, rather than just the moves that occur in the game, and could help give insights into their motivation.
15
The experiments were designed to test the following theoretical hypotheses:
Hypothesis 1: Compared to the individual lending (treatment 1), the lending rate, the average
level of monitoring and the average repayment rate are all higher in the group lending
treatments (treatments 2 and 3).
Hypothesis 2: Compared to the sequential group lending treatment (treatment 3), the lending
rate, the average level of monitoring and the average repayment rate are not higher in the
simultaneous group lending treatment (treatment 2).
Note that the weak inequalities indicated in Hypothesis 2 follow from the theoretical
predictions that the efficient (lend/monitor) equilibrium is unique in the sequential lending
treatment, but both efficient and inefficient (no loan) equilibria exist in the simultaneous
lending case.
5. Results
We present our results in the next four subsections, with each subsection dealing with a
specific aspect of the program: lending, monitoring, repayment (and project choice) and
finally efficiency. In each case we present conservative non-parametric tests for treatment
differences which require minimal statistical assumptions and are based on only one
independent summary statistic value per session. We also report estimates from multivariate
parametric regression models which can isolate the contribution of different factors on lender
and borrower behavior.
5.1: Lending
Figure 2 presents the average proportion of lenders making loans in the different periods, by
treatment. Clearly the average proportion of lenders making loans is substantially lower at
every period for treatment 1 (individual lending) but there is very little difference between
treatments 2 and 3 (group lending), providing initial support for both hypothesis 1 and 2.
These results are supported by non-parametric Mann-Whitney rank sum tests with the session
16
average as the unit of observation (see Table 3, Panel A). Compared to the initial 5 periods,
the average proportion making loans is also significantly lower in the later 5 periods in
treatment 1 (Wilcoxon signed rank test 0 for 8; 0.01nS n p value= = − < ), but is not
significantly different between early and late periods in treatment 2
( )12.5 for 8; not significantnS n= = and treatment 3 ( )10.5 for 8; not significantnS n= = .
Subjects participated in the experiment over 40 periods, allowing us to examine their
behavior over time more systematically using panel regressions. Table 4 presents the random
effect probit estimation of the lender making the loan. These panel regressions incorporate a
random effects error structure, with the subject (lender) representing the random effect. The
dependent variable is the propensity to lend. The explanatory variables that we include are:
dummies for simultaneous and sequential lending treatment (the reference category is
individual lending); a dummy variable indicating whether the lender made a loss in the
previous period; a dummy for the session being run at Jadavpur University in Kolkata (to
capture subject differences across the two countries); and two variables that capture the effect
of time on the propensity to make the loan in a particular period: 1t and the interaction term
( ) ( )1 GROUPt × , where GROUP is a dummy variable that takes the value of 1 if group
lending treatment (simultaneous and sequential lending) and 0 if individual lending
treatment.15 We also include a set of demographic controls.
15 Notice from Figure 2 that the time trend appears quite similar for the two group lending treatments but is very
different for the individual lending treatment. The non-interacted term ( )1t in this case captures the effect of
time on the propensity of the lender to make a loan in the individual lending treatment while the interaction term captures the differential effect of time on the propensity of the lender to make a loan in the group lending treatment. To get the total effect of time in the group lending treatments we need to add the coefficient estimates
of ( )1t and ( ) ( )( )1 GROUPt × .
17
Table 4 presents the coefficient estimates and standard errors. Lending decreased over
time in treatment 1, but increased over time in the two group lending treatments.16 The two
treatment dummies are both positive and statistically significant indicating that the
probability of lending is significantly higher in group lending compared to individual lending.
This provides further support for Hypothesis 1. The difference in the probability of lending in
the two group lending treatments is however statistically significant at the 10% level (using
the test of equality across the two group lending treatment dummies), weakly contradicting
Hypothesis 2. The probability of lending in period t is significantly lower if the lender
received negative earnings in period 1t − . The Jadavpur University dummy is not statistically
significant implying that that there is no difference in the probability of lending across the
two samples. Most of the demographic control variables are not statistically significant.
5.2: Monitoring
Figure 3 presents the average level of monitoring across periods. With the exception of the
first few periods, average monitoring is highest in the sequential lending treatment and lowest
in the individual lending treatment. Again using a rank sum test with the session average as
the unit of observation, the difference in monitoring rate between treatments 2 and 3 is not
statistically significant (Table 3, Panel B). The average monitoring rate is always
significantly lower for treatment 1, however. This provides support for Hypotheses 1 and 2.
In treatment 1 (individual lending) the monitoring rate is lower in the last 5 periods than in
the first 5 periods (Wilcoxon signed rank test 2 for 8; 0.05nS n p value= = − < ), indicating a
significant decline in monitoring rates. By contrast, monitoring rates increase between the
first 5 and last 5 periods for treatment 2 (Wilcoxon 4 for 8; 0.05nS n p value= = − < ) and
treatment 3 (Wilcoxon 0 for 8; 0.01nS n p value= = − < ).
16 The coefficient estimates of ( )1
t and ( ) ( )( )1 GROUPt × are jointly significant.
18
The monitoring decision is made by different agents in the individual and group
lending treatments. Hence we analyze the level of monitoring chosen in the individual and
group lending treatments separately.17 The level of monitoring chosen is restricted in the
range ( )0,1 and is estimated using a tobit model.
Consider first the level of monitoring chosen (by the lender) in the individual lending
treatment. The explanatory variables are the same as those in Table 4, except that here, by
definition, we do not include the interaction term ( ) ( )1 GROUPt × . Table 5, Panel A,
presents the tobit regression results with player fixed effects and the Hausman-Taylor
estimates for error component models.18 The level of monitoring in period 1t − has a positive
and statistically significant impact on the level of monitoring in period t . The Jadavpur
University dummy is not statistically significant, indicating no locational differences.
As mentioned above in the case of group lending (with peer monitoring) the payoff
for an individual borrower depends both on her level of monitoring and also on the level of
monitoring of her partner. Subjects could construct expectations for the level of monitoring of
the other member of the group in different ways. Here we consider the following two simple
alternatives:
(1) Cournot expectations: each subject expects the monitoring level of the other member of the group to be the same as that in the previous period (Lagged Other Monitoring);
(2) Fictitious play: each subject expects the monitoring level of the other member of the group to be the average over all the previous periods (Lagged Average Other
17 The propensity to make the loan is significantly lower in the individual lending treatment, implying that the data on the level of monitoring is often not observed in the individual lending treatment. The panel in this case is therefore unbalanced: the observed number of monitoring choices varies from 2 (in only 2 of the possible 40 cases, did they choose to make the loan) to 37. 18 The tobit regression results presented in column (1) fail to account for the possibility that the lagged dependent variable (lagged level of monitoring) can be correlated with the time invariant component of the error term (the unobserved individual level random effect). Ignoring this could result in biased estimates. One way of obtaining unbiased estimates would be to use instrumental variables estimation (see Hausman and Taylor, 1981). It is assumed that none of the covariates is correlated with the idiosyncratic error term. The results for the Hausman-Taylor estimation for error component models are presented in Table 5, Panel A, column (2). Qualitatively the results are very similar to the tobit regression results presented in column (1): in particular, the greater the level of monitoring in period 1t − , the greater the level of monitoring in period t and the level of monitoring falls over time.
19
Monitoring). Hence each subject is assumed to have a long memory as opposed to the Cournot expectations case where each subject has a short memory.
Table 5, Panel B, presents the random effects tobit and the Hausman-Taylor estimation for
error component models for both specifications of expectation formation in the group lending
treatment. We find that monitoring increased over time. The positive and significant
coefficient estimate of the other borrower’s lagged monitoring level (in the Cournot
expectations version) or its counterpart lagged average other borrower’s monitoring (in the
fictitious play version) is consistent with the upwardly-sloped reaction functions of the
theoretical model. Note that the coefficient estimate on a borrower’s own monitoring in the
previous period is also positive, and is substantially larger than the reaction to the other
borrower’s monitoring level. The sequential lending treatment dummy is never statistically
significant, consistent with Hypothesis 2. The Jadavpur University dummy is always negative
but is statistically significant only in the tobit regressions implying that subjects in India
choose a lower level of peer monitoring.19 Finally turning to the demographic controls, we
find that females choose a significantly lower level of monitoring, as do subjects with no
previous participation experience.
5.3: Repayment Rate
Figure 4 presents the average repayment rates in the three treatments by period. Note that this
repayment rate is not a choice variable but is the result of a combination of the ex ante project
choice by the borrower, the level of monitoring chosen by the borrower, and the success of
the monitoring process. Repayment occurs if the borrower chooses project S or if the
borrower chooses project R and monitoring is successful. It is clear from Figure 4 that
average repayment rates are lowest in the individual lending treatment. The results are again
supported by a rank sum test with the session average as the unit of observation (Table 3,
19 It is interesting to note that if we restrict the sample to those born in South Asia (whether residing in Australia or India), the Jadavpur University dummy is no longer statistically significant.
20
Panel C). The high repayment rates in the group lending treatments are essentially driven by
the high monitoring rates and not by the borrowers ex ante choosing project S. Panel D of
Table 3 presents the average proportion of borrowers choosing project R ex ante. The average
proportion of subjects (ex ante) choosing project R is significantly lower in the individual
lending treatment. Note also that the proportion of borrowers choosing project R is not
different across treatments in the initial periods but becomes significantly different across all
treatments by the later periods.
Recall that the earnings of the borrower are greater if he chooses project R, but the
earnings of the lender are lower if the borrowers choose project R. In treatments 2 and 3, the
borrowers are more likely to choose project R, but they are also more likely to choose a high
level of monitoring to be able to switch the other borrower’s project choice to S. In
consequence the “actual project choices” are likely to be project S and the earnings of the
lenders are positive and we move to an efficient (monitoring/lending) equilibrium. On the
other hand in treatment 1 monitoring rates are lower (monitoring is also more costly ( )1λ > )
and even though borrowers are more likely to choose project S, lenders choose not to make
the loan. We end up at the inefficient (low monitoring/no lending) equilibrium. The results
therefore imply that group lending with peer monitoring works better compared to individual
lending with lender monitoring.
Table 6 presents random effect probit regression results for repayment and choice of
project R. The explanatory variables are the same as in Table 4. The repayment rates (Table
6, column 1) are significantly higher in the group lending treatments compared to the
individual lending treatment, consistent with Hypothesis 1. The sequential lending treatment
has significantly higher repayment rates compared to the simultaneous lending treatment (the
test of equality of treatment effects is significant, indicating that the two group lending
treatments do not have similar effects on repayment), consistent with Hypothesis 2. The
21
Jadavpur University dummy is not statistically significant and finally repayment rates are
significantly lower for subjects who had a higher proportion of correct answers in the quiz.
Column 2 presents the random effects probit regression results for choice of project R.
The probability of choosing project R decreased over time in treatment 1, while the
probability of choosing project R increased over time in the two group lending treatments.20
The two treatment dummies are both positive and statistically significant, indicating that the
probability of choosing project R is significantly higher in group lending compared to
individual lending. The probability of choosing project R is no different in the two group
lending treatments. In all three treatments, subjects who seek to maximize their own current
period monetary earnings should always choose project R. We observe, however, that
borrowers have a lower probability of choosing R in the individual lending treatment. One
possible explanation is that reciprocal motivations are triggered more in a two person game
than a three person game.
5.4: Efficiency
Finally we turn to efficiency, defined as the total surplus attained by the borrowers and the
lenders as a proportion of the maximum possible surplus. Here efficiency is used as a
measure of the overall performance of the market. Maximum surplus is attained when the
lender provides the loan; the level of monitoring chosen is 0 and the borrowers choose project
S. Panel E of Table 3 indicates that average efficiency is lowest in the individual lending
treatment. The higher monitoring rates in the sequential lending treatment could lead to lower
efficiency than the simultaneous group lending treatment, but the differences are only
significant in the early periods.
6. Concluding Comments
20 Again the coefficient estimates of ( )1
t and ( ) ( )( )1 GROUPt × are jointly significant.
22
This paper reports the results from a laboratory experiment of group lending in the presence
of moral hazard and costly lender or peer monitoring. We compare treatments when credit is
provided to members of the same group sequentially and simultaneously, and when loans are
given to individuals and monitored by lenders. Our results suggest that in the presence of
moral hazard, peer monitoring results in higher loan frequencies, higher monitoring and
higher repayment rates compared to bank monitoring. The differences are minor between
simultaneous and sequential lending.
Results from this experiment can help shed light on the important policy issue of the
optimal design for microcredit programs. Much of the success of microcredit programs has
been attributed to self-selected groups and social ties in rural communities. However
successful application of these programs in other scenarios and economies requires more than
strong social ties. In urban contexts of developing and transitional economies, for example, it
might be more difficult to form self-selected borrowing groups compared to the more closely
knit rural communities. For this reason several authors and policy makers suggest that
optimal design of microcredit programs look beyond the issue of self-selection and even look
beyond group lending. In this experiment, we focus on informational asymmetries due to
moral hazard and we restrict ourselves to exogenously formed groups. Our results show that
in the presence of moral hazard group lending performs better compared to individual
lending, even with no self-selection in group formation. Introducing dynamic incentives
(within group lending) helps, but not significantly. What is important is peer monitoring,
which works much better than active lender monitoring.21 Optimal design of microcredit
programs needs to take advantage of the fact that it is less costly for group members to
monitor each other, which can result in better project choices and higher repayment rates.
21 It has been observed that in the absence of peer monitoring the success of such programs is quite limited. See Bhatt and Tang (2002) for evidence using data from microcredit programs in the US..
References: Abbink, K., B. Irlenbusch and E. Renner (2006): “Group Size and Social Ties in
Microfinance Institutions”, Economic Inquiry, 44(4), 614 – 628. Aniket, K. (2006): “Sequential Group Lending with Moral Hazard”, Mimeo London School of
Economics. Armendariz de Aghion, B. and C. Gollier (2000): “Peer Group Formation in an Adverse
Selection Model”, Economic Journal, 110, 632 – 643. Armendariz de Aghion, B. and J. Morduch (2000): “Microfinance beyond Group Lending”
Economics of Transition, 8, 401 – 420. Armendariz de Aghion, B. and J. Morduch (2005): The Economics of Microfinance, MIT
Press. Banerjee, A. V., T. Besley and T. W. Guinnane (1994): “Thy Neighbor’s Keeper: The Design
of a Credit Cooperative with Theory and a Test”, Quarterly Journal of Economics, 109, 491 – 515.
Barr, A (2001): “Social Dilemmas and Shame-Based Sanctions: Experimental Results from Rural Zimbabwe”, Mimeo, Centre for the Study of African Economies, University of Oxford.
Bellman, E. (2006): “Invisible Hand: Entrepreneur Gets Big Banks to Back Very Small Loans; Microlending-for-Profit Effort In India Draws Business From Citigroup, HSBC; Ms. Dobbala's Baby Buffalo”, Wall Street Journal, Eastern Edition, 15/5/2006, A1.
Bhatt, N. and S-Y. Tang (2002): “Determinants of Repayment in Microcredit: Evidence from Programs in the United States”, International Journal of Urban and Regional Research, 26(2), 360 – 370.
Bolnik, B. R. (1988): “Evaluating Loan Collection Performance: An Indonesian Example”, World Development, 16, 501 – 510.
Carpenter, J., S. Bowles and H. Gintis (2006): “Mutual Monitoring in Teams: Theory and Experimental Evidence on the Importance of Reciprocity”, IZA Discussion Paper # 2106.
Cassar, A., L. Crowley and B. Wydick (2007): “The Effect of Social Capital on Group Loan Repayment: Evidence from Field Experiments”, Economic Journal, forthcoming.
Che Y-K. (2002): “Joint Liability and Peer Monitoring under Group Lending”, Contributions to Theoretical Economics, 2(1).
Chowdhury, P. R. (2005): “Group-lending: Sequential Financing, Lender Monitoring and Joint Liability”, Journal of Development Economics, 77, 415 – 439.
Conyabuguma, M., T. Page and L. Putterman (2005): “Cooperation under the Threat of Expulsion in a Public Goods Experiment”, Journal of Public Economics, 89(8), 1421 – 1435.
Conlin, M. (1999): “Peer Group Micro-Lending Programs in Canada and the United States”, Journal of Development Economics, 60, 249 – 269.
Fehr, E. and S. Gaechter (2000): “Cooperation and Punishment in Public Goods Experiments”, American Economic Review, 90(4), 980 – 994.
Fischbacher, U. (2007): “z-Tree-Zurich Toolbox for Readymade Economic Experiments,” Experimental Economics, forthcoming.
Fry, T. R. L., S. Mihajilo, R. Russell and R. Brooks (2006): “The Factors Influencing Saving in a Matched Savings Program: The Case of the Australian Saver Plus Program”, Mimeo, RMIT University, Melbourne.
Ghatak, M. and T. W. Guinnane (1999): “The Economics of Lending with Joint Liability”, Journal of Development Economics, 60, 195 – 228.
Ghatak, M. (2000): “Screening by the Company You Keep: Joint Liability Lending and the Peer Selection Effect”, Economic Journal, 110, 601 – 631.
24
Gine, X., P. Jakiela, D. S. Karlan and J. Morduch (2005): “Microfinance Games”, Mimeo Yale University.
Gine, X. and D. S. Karlan (2006): “Group versus Individual Liability: A Field Experiment in the Philippines”, Mimeo, Yale University.
Greif, A. (1994): “Cultural Beliefs and the Organization of Society: A Historical and Theoretical Reflection on Collectivist and Individualist Societies”, Journal of Political Economy, 102(5), 912 – 950.
Hausman, J. A. and W. E. Taylor (1981): “Panel Data and Unobservable Individual Effects”, Econometrica, 49, 1377 – 1398.
Hulme, D. (2000): “Impact Assessment Methodologies for Microfinance: Theory and Better Practice”, World Development, 28, 79 – 98.
Kono, H. (2006): “Is Group Lending a Good Enforcement Scheme for Achieving High Repayment Rates? Evidence from Field Experiments in Vietnam”, Mimeo, Institute of Developing Economies, Chiba, Japan.
Masclet, D., C. Noussair, S. Tucker and M.C. Villeval (2003): “Monetary and Nonmonetary Punishment in the Voluntary Contributions Mechanism”, American Economic Review, 93(1), 366 – 380.
Morduch, J. (1999): “The Microfinance Promise”, Journal of Economic Literature, 37, 1564 – 1614.
Pulley, R (1989): “Making the Poor Creditworthy: a Case Study of the Integrated Rural Development Program in India“, World Bank Discussion Paper 58, Washington DC: World Bank.
Rai, A. and T. Sjostrom (2004): “Is Grameen Lending Efficient? Repayment Incentives and Insurance in Village Economies”, Review of Economic Studies, 71 (1), 217 – 234.
Ray, D. (1998): Development Economics, Princeton University Press. Schelling, T. (1980): The Strategy of Conflict (Cambridge: Harvard University Press). Seddiki, M. W. and M. Ayedi (2005): “Cooperation and Punishment in Group Lending, the
Experimental Case”, Mimeo. Stiglitz, J. (1990): “Peer Monitoring and Credit Markets”, World Bank Economic Review, 4,
351 – 366. Van Tassel, E. (1999): “Group Lending Under Asymmetric Information”, Journal of
Development Economics, 60, 3 – 25. Van Huyck, J., R. Battalio and R. Beil (1990): “Tacit Coordination Games, Strategic
Uncertainty, and Coordination Failure,” American Economic Review, 80, 234 – 248. Varian, H. (1990): “Monitoring Agents with Other Agents”, Journal of Institutional and
Theoretical Economics, 146, 153 – 174. Walker, J. and M. Halloran (2004): “Rewards and Sanctions and the Provision of Public
Goods in One-Shot Settings”, Experimental Economics, 7(3), 235 – 247.
25
Table 1: Relationship between C and M Panel A: Treatments 2 and 3 (Group Lending, Peer Monitoring) Monitoring Cost (C)
M Interpretation of M percentage:
$0.000 0% Switch a borrower choice of R to S 0 out of 10 times $0.005 10% Switch a borrower choice of R to S 1 out of 10 times on average $0.020 20% Switch a borrower choice of R to S 2 out of 10 times on average $0.045 30% Switch a borrower choice of R to S 3 out of 10 times on average $0.080 40% Switch a borrower choice of R to S 4 out of 10 times on average $0.125 50% Switch a borrower choice of R to S 5 out of 10 times on average $0.180 60% Switch a borrower choice of R to S 6 out of 10 times on average $0.245 70% Switch a borrower choice of R to S 7 out of 10 times on average $0.320 80% Switch a borrower choice of R to S 8 out of 10 times on average $0.405 90% Switch a borrower choice of R to S 9 out of 10 times on average $0.500 100% Switch a borrower choice of R to S 10 out of 10 times Panel B: Treatment 1 (Individual Lending, Bank Monitoring) Monitoring Cost (C)
M Interpretation of M percentage:
0.0000 0% Switch a borrower choice of R to S 0 out of 10 times 0.0225 10% Switch a borrower choice of R to S 1 out of 10 times on average 0.0900 20% Switch a borrower choice of R to S 2 out of 10 times on average 0.2025 30% Switch a borrower choice of R to S 3 out of 10 times on average 0.3600 40% Switch a borrower choice of R to S 4 out of 10 times on average 0.5625 50% Switch a borrower choice of R to S 5 out of 10 times on average 0.8100 60% Switch a borrower choice of R to S 6 out of 10 times on average 1.1025 70% Switch a borrower choice of R to S 7 out of 10 times on average 1.4400 80% Switch a borrower choice of R to S 8 out of 10 times on average 1.8225 90% Switch a borrower choice of R to S 9 out of 10 times on average 2.2500 100% Switch a borrower choice of R to S 10 out of 10 times
26
Table 2: Earnings of Borrowers and Lenders Panel A: Treatment 2 (Simultaneous Lending)
Actual project choice of
borrower 1
Actual project choice of
borrower 2
Earnings of borrower 1
Earnings of borrower 2
Earnings of lender
S S $1.75 – C1 $1.75 – C2 $2.50 S R $0.00 – C1 $2.50 – C2 $2.00 R S $2.50 – C1 $0.00 – C2 $2.00 R R $2.50 – C1 $2.50 – C2 -$2.00
No loan is provided $0.00 $0.00 $1.50 Note: C1 and C2 denote the monitoring costs incurred by borrower 1 and 2. Panel B: Treatment 3 (Sequential Lending)
Actual project choice of the first
borrower
Actual project choice of the
second borrower
Earnings of first borrower
Earnings of second borrower
Earnings of lender
S S $1.75 – C1 $1.75 – C2 $2.50 S R $0.00 – C1 $2.50 – C2 $2.00 R S $2.50 – C1 $0.00 – C2 -$0.25 R R $2.50 – C1 $0.00 – C2 -$0.25
No loan is provided $0.00 $0.00 $1.50 Note: C1 and C2 denote the monitoring costs incurred by the first and the second borrower. Panel C: Treatment 1 (Individual Lending)
Actual project choice of borrower
Earnings of borrower Earnings of lender
S $1.75 $1.25 – C R $2.50 –$1.00 – C
No loan is provided $0.00 $0.75 Note: C denotes the monitoring cost incurred by the lender
27
Table 3: Selected Descriptive Statistics Panel A. Average Proportion Making Loans Full Sample First 5 periods Last 5 periods Individual Lending Treatment (Treatment 1)
0.4720 0.5875 0.2938
Simultaneous Lending Treatment (Treatment 2)
0.8138 0.7563 0.8047
Sequential Lending Treatment (Treatment 3)
0.7295 0.7000 0.8732
Rank sum Test Treatment 1 = Treatment 2 -3.256*** -2.107** -3.376*** Treatment 1 = Treatment 3 -2.102*** -1.263 -3.258*** Treatment 2 = Treatment 3 0.684 0.582 -0.820 Individual Lending = Group Lending -3.124*** -1.965** -3.890*** Panel B. Average Level of Monitoring Full Sample First 5 periods Last 5 periods Individual Lending Treatment (Treatment 1)
0.3268 0.4142 0.2466
Simultaneous Lending Treatment (Treatment 2)
0.5658 0.5094 0.6442
Sequential Lending Treatment (Treatment 3)
0.6178 0.5038 0.7213
Rank sum Test Treatment 1 = Treatment 2 -3.009*** -0.927 -3.009*** Treatment 1 = Treatment 3 -3.125*** -1.852* -3.134*** Treatment 2 = Treatment 3 -0.840 -0.210 -0.811 Individual Lending = Group Lending -3.541*** -1.604 -3.561*** Panel C. Average Repayment Rates Full Sample First 5 periods Last 5 periods Individual Lending Treatment (Treatment 1)
0.4756 0.5000 0.3938
Simultaneous Lending Treatment (Treatment 2)
0.6555 0.6156 0.7242
Sequential Lending Treatment (Treatment 3)
0.6979 0.6500 0.7571
Rank sum Test Treatment 1 = Treatment 2 -2.836** -1.631 -2.897*** Treatment 1 = Treatment 3 -3.361*** -2.209** -3.258*** Treatment 2 = Treatment 3 -0.420 -0.423 0.0000 Individual Lending = Group Lending -3.613*** -2.241** -3.593***
28
Panel D. Average Proportion Choosing the Non-Verifiable Project R Full Sample First 5 periods Last 5 periods Individual Lending Treatment (Treatment 1)
0.6293 0.6667 0.6646
Simultaneous Lending Treatment (Treatment 2)
0.7999 0.7219 0.8750
Sequential Lending Treatment (Treatment 3)
0.7951 0.6906 0.8643
Rank sum Test Treatment 1 = Treatment 2 -2.626*** -0.632 -2.370** Treatment 1 = Treatment 3 -2.836*** -0.528 -2.437** Treatment 2 = Treatment 3 0.053 0.053 0.582 Individual Lending = Group Lending -3.185*** -0.675 -2.816*** Panel E. Efficiency Full Sample First 5 periods Last 5 periods Individual Lending Treatment (Treatment 1)
0.4470 0.4841 0.3603
Simultaneous Lending Treatment (Treatment 2)
0.6616 0.6300 0.6715
Sequential Lending Treatment (Treatment 3)
0.5630 0.5389 0. 6547
Rank sum Test Treatment 1 = Treatment 2 -3.361*** -2.731*** -3.151*** Treatment 1 = Treatment 3 -1.890*** -1.050** -3.125*** Treatment 2 = Treatment 3 1.470 1.890* 0.231 Individual Lending = Group Lending -3.062*** -2.205*** -3.679***
29
Table 4: Random Effect Probit Regression for Making Loans Coefficient Estimate 1/t 1.9262*** (0.3471) 1/t × GROUP -2.4518*** (0.4917) Simultaneous Lending Treatment (Dummy) 1.3629*** (0.2116) Sequential Lending Treatment (Dummy) 0.9684*** (0.2127) Negative Earnings in Previous Period (Dummy) -0.4416*** (0.0593) Session at Jadavpur University (Dummy) -0.2914 (0.1782) Proportion of Correct Answer in Quiz -0.0977 (0.6236) Age of Subject -0.2953 (0.1925) Square of Age of Subject 0.0051 (0.0037) Female 0.2759 (0.1759) Business/Economics/Commerce Major 0.1195 (0.1704) Lived in a Suburb at Age 15 -0.0324 (0.2514) Lived in a Large City at Age 15 -0.1969 (0.3153) Lived in a Metropolis at Age 15 0.3811* (0.2093) 2nd Year -0.0882 (0.2063) 3rd Year 0.2710 (0.2167) No Previous Participation in Experiments -0.3771 (0.2428) Constant 3.9590 (2.5991) Observations 4112 Number of group(session subject) 108 σu 0.7424*** (0.0626) ρ 0.3553*** (0.0387) LR Test for ρ = 0 624.63*** Equality of Treatment Effects 2.93* Standard errors in parentheses * significant at 10%; ** significant at 5%; *** significant at 1%
30
Table 5: Level of Monitoring Chosen. Panel A: Individual Lending (Lender Monitoring) Tobit Regression with
Player Fixed Effects Hausman-Taylor
Estimation for Error Component Models
1/t 0.1861** 0.1752** (0.0916) (0.0870) Lagged Monitoring 0.2848*** 0.2449*** (0.0515) (0.0483) Session at Jadavpur University (Dummy) -0.0670 0.0549 (0.0821) (0.0653) Proportion of Correct Answer in Quiz -1.6822*** -1.4880 (0.5033) (1.2639) Age of Subject 0.1043 -0.2617 (0.6128) (0.2877) Square of Age of Subject -0.0037 0.0049 (0.0139) (0.0052) Female 0.1960 0.1519** (0.1811) (0.0718) Business/Economics/Commerce Major 0.1785** 0.0429 (0.0778) (0.0848) Lived in a Suburb at Age 15 0.3329 0.0716 (0.2091) (0.0749) Lived in a Large City at Age 15 0.1480 -0.0399 (0.1938) (0.0828) Lived in a Metropolis at Age 15 0.3681*** 0.0756 (0.1381) (0.0614) 2nd Year -0.4169* -0.0979 (0.2337) (0.0848) 3rd Year -0.3329*** 0.0222 (0.1275) (0.1087) No Previous Participation in Experiments -0.1483 -0.0929 (0.1016) (0.1250) Constant 1.0901 4.7687 (6.7142) (4.8930) Observations 477 477 Number of group(session subject) 41 σ 0.2077*** (0.0075) σu 0.0979 σe 0.1892 ρ 0.2111 Joint Significance of Player Fixed Effects 3.00*** Left censored observations 64 Uncensored observations 406 Right censored observations 7 Standard errors in parentheses * significant at 10%; ** significant at 5%; *** significant at 1%; ρ denotes the fraction of total variance due to the time invariant component of the error term.
31
Table 5 (Panel B): Group Lending (Peer Monitoring) Cournot Play Fictitious Play Random
Effects Tobit Regression
Hausman-Taylor
Estimation for Error
Component Models
Random Effects Tobit Regression
Hausman-Taylor
Estimation for Error
Component Models
1/t -0.1294*** -0.1559*** -0.1162** -0.1414*** (0.0497) (0.0381) (0.0503) (0.0386)
0.0298 0.0277 0.0288 0.0257 Sequential Lending Treatment (Dummy) (0.0264) (0.0210) (0.0265) (0.0211) Lagged Own Monitoring 0.5149*** 0.3616*** 0.5026*** 0.3493*** (0.0192) (0.0140) (0.0194) (0.0142)
0.1149*** 0.0932*** Lagged Monitoring of the Other Borrower (0.0164) (0.0125)
0.2952*** 0.2619*** Average Lagged Monitoring of the Other Borrower (0.0507) (0.0390)
-0.0649** -0.0349 -0.0575* -0.0289 Session at Jadavpur University (Dummy) (0.0309) (0.0255) (0.0306) (0.0256)
-0.0608 -0.0561 -0.0759 -0.0646 Proportion of Correct Answer in Quiz (0.0776) (0.0577) (0.0770) (0.0581) Age of Subject 0.0761 -0.1673 0.0749 -0.1902 (0.0858) (0.1553) (0.0913) (0.1542) Square of Age of Subject -0.0020 0.0036 -0.0019 0.0042 (0.0020) (0.0036) (0.0021) (0.0035) Female -0.0768*** -0.0472** -0.0898*** -0.0588*** (0.0257) (0.0198) (0.0257) (0.0201)
0.0051 -0.0042 -0.0037 -0.0105 Business/Economics/Commerce Major (0.0262) (0.0204) (0.0258) (0.0206) Lived in a Suburb at Age 15 -0.0558 -0.0681** -0.0516 -0.0683** (0.0391) (0.0314) (0.0391) (0.0316) Lived in a Large City at Age 15 -0.0041 -0.0064 0.0015 -0.0011 (0.0373) (0.0299) (0.0385) (0.0301) Lived in a Metropolis at Age 15 -0.0031 -0.0100 -0.0021 -0.0088 (0.0316) (0.0243) (0.0309) (0.0245) 2nd Year -0.0039 0.0083 0.0001 0.0174 (0.0312) (0.0245) (0.0317) (0.0247) 3rd Year -0.0123 0.0076 -0.0096 0.0136 (0.0353) (0.0316) (0.0368) (0.0318)
-0.0822** -0.0953*** -0.0858** -0.1000*** No Previous Participation in Experiments (0.0386) (0.0315) (0.0386) (0.0317) Constant -0.2837 2.4053 -0.3530 2.5714 (0.9138) (1.6854) (0.9749) (1.6751) Observations 4504 4504 4504 4504 Number of group(session subject)
120 120 120 120
σu 0.1186*** 0.0934 0.1206*** 0.0942 (0.0101) (0.0099) σe 0.3037*** 0.2395 0.3041*** 0.2398 (0.0040) (0.0040) ρ 0.1322*** 0.1319 0.1358*** 0.1336 (0.0198) (0.0195) LR Test for σu = 0 230.80*** 247.60***
32
Left censored observations 444 444 Uncensored observations 3360 3360 Right censored observations 700 700 Standard errors in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% ρ denotes the fraction of total variance due to the time invariant component of the error term.
33
Table 6: Random Effect Probit Regressions for Repayment and Choice of Non-Verifiable Project (R) Repayment Non-Verifiable Project
Choice 1/t -0.0293 0.3701** (0.1719) (0.1883) 1/t × GROUP -0.2123 -1.3024*** (0.2047) (0.2244) Simultaneous Lending Treatment (Dummy) 0.4639*** 0.8871*** (0.0892) (0.1590) Sequential Lending Treatment (Dummy) 0.6241*** 0.7471*** (0.0871) (0.1537) Session at Jadavpur University -0.0473 -0.2414* (0.0799) (0.1428) Proportion of Correct Answer in Quiz -0.4500** 1.3297*** (0.2111) (0.3826) Age of Subject -0.0189 0.1223 (0.2461) (0.4409) Square of Age of Subject 0.0007 -0.0037 (0.0057) (0.0102) Female 0.1173* -0.2287* (0.0700) (0.1274) Business/Economics/Commerce Major 0.0046 0.0491 (0.0734) (0.1339) Lived in a Suburb at Age 15 -0.0226 -0.1576 (0.1045) (0.1880) Lived in a Large City at Age 15 -0.1513 0.1259 (0.1065) (0.1963) Lived in a Metropolis at Age 15 0.0050 -0.0798 (0.0867) (0.1579) 2nd Year -0.2072** 0.1935 (0.0848) (0.1557) 3rd Year -0.1491 0.1749 (0.0973) (0.1763) No Previous Participation in Experiments 0.0901 -0.2950 (0.1060) (0.1956) Constant 0.4057 -1.2075 (2.6438) (4.7432) Observations 6532 6532 Number of group(session subject) 168 168 σu 0.3734*** 0.7226*** (0.0280) (0.0510) ρ 0.1224*** 0.3430*** (0.0161) (0.0318) LR Test for ρ = 0 259.23*** 825.31*** Equality of Treatment Effects 3.83** 0.87 Standard errors in parentheses * significant at 10%; ** significant at 5%; *** significant at 1%
34
Figure 1: Reaction Functions in Simultaneous lending. Note that reaction functions intersect in two places (at (0, 0) and at (1, 1)), which leads to multiple equilibria. (0, 0) (1, 1)
Reaction functions for (H-r)=1.75
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
mi
mj
j's reaction functioni's reaction function
35
Figure 2: Average Proportion Making Loan, by treatment
0.1
.2.3
.4.5
.6.7
.8.9
1A
vera
ge P
ropo
rtion
Mak
ing
Loan
0 5 10 15 20 25 30 35 40Period
Individual Lending Simultaneous Group LendingSequential Group Lending
36
Figure 3: Average Monitoring Level, by Treatment
0.1
.2.3
.4.5
.6.7
.8.9
1A
vera
ge M
onito
ring
Leve
l
0 5 10 15 20 25 30 35 40Period
Individual Lending Simultaneous Group LendingSequential Group Lending
37
Figure 4: Repayment Rate, by Treatment
0.1
.2.3
.4.5
.6.7
.8.9
1A
vera
ge R
epay
men
t Rat
e
0 5 10 15 20 25 30 35 40Period
Individual Lending Simultaneous Group LendingSequential Group Lending
38
T2 - L - 200306 Instructions
General: This is an experiment in the economics of decision-making. The instructions are simple and if you follow them carefully and make good decisions you will earn money that will be paid to you privately in cash at the end of the experimental session. Your earnings will be in experimental dollars and they will be converted into real dollars at the following rate: 1 Experimental Dollar = ____ Real Dollars. Notice that you earn more money by earning more experimental dollars. After we finish reading the instructions and before we start the experiment, we would like you to answer a set of questions relating to these instructions. You will be paid in cash (at the end of the experiment, in addition to your earnings from the actual experiment) at the rate of $0.50 for each correct answer. In today’s experiment, you will be randomly divided into groups and each group will have three members. Each group consists of one lender and two borrowers. Your role—either borrower or lender—is determined randomly and will remain unchanged throughout the experiment. At the end of every period, participants will be randomly re-matched and so the other people in your group will typically change each period. You will make decisions for 40 periods. Decision Making: Two projects are available to each borrower every period: project S and project R. The cost of each project is $1 and it is to be financed by a loan from the lender. Every period the lender can choose whether or not to invest her $2 into making loans to the borrowers. She must either make the loan to both borrowers or to neither borrower, and she cannot make a loan to a single borrower. If the lender chooses not to invest in the loans to the borrowers, she earns $1.50 for the period. If the borrowers receive the loans, they can monitor the project choice of the other borrower in their group by choosing to pay a monitoring cost (C). Both borrowers can monitor each other. If borrower X incurs a cost C on monitoring, there is a chance of M that the other borrower Y will automatically be required to choose project S. Otherwise the other borrower can choose either project S or project R. Choices will be made simultaneously and the borrowers will not know whether the lender chooses to make the loans or not before making their choice of project. All decisions will be revealed after both the lender and the borrowers have made their decisions. Borrowers pay their selected monitoring costs whenever the lender makes the loan, regardless of whether or not the monitoring is successful. The monitoring chances work in the following way. Suppose borrower X chooses M = 20%. In this case, imagine an urn (or the bingo cage the experimenter is holding) containing 10 total balls: 2 white balls and 8 red balls. One ball is drawn from this
39
imaginary urn, and if we draw a white ball then a borrower Y choice of R is switched to S; if we draw a red ball then a borrower Y choice of R remains R. If borrower Y chose S, then this choice of S is implemented regardless of the ball draw. Remember, borrower Y also makes monitoring choices in the same way to possibly switch borrower X choices from R to S. To take another example, suppose borrower Y chooses M = 70%. In this case you should imagine an urn containing 7 white balls and 3 red balls. Again, a drawn white ball switches a borrower X choice of R to S, but a drawn red ball means that a borrower X choice of R remains R. Therefore, a higher choice of M, which is more costly as shown in the table below, increases the chances that the other borrower’s choice of R is switched to S. A different ball draw, from a different imaginary urn, is conducted for every different group and borrower for every different period in the experiment. In other words, the random draws are all independent. The relationship between C and M is as follows: Monitoring Cost (C)
M
Interpretation of M percentage:
$0.000 0% Switch a borrower choice of R to S 0 out of 10 times $0.005 10% Switch a borrower choice of R to S 1 out of 10 times on average $0.020 20% Switch a borrower choice of R to S 2 out of 10 times on average $0.045 30% Switch a borrower choice of R to S 3 out of 10 times on average $0.080 40% Switch a borrower choice of R to S 4 out of 10 times on average $0.125 50% Switch a borrower choice of R to S 5 out of 10 times on average $0.180 60% Switch a borrower choice of R to S 6 out of 10 times on average $0.245 70% Switch a borrower choice of R to S 7 out of 10 times on average $0.320 80% Switch a borrower choice of R to S 8 out of 10 times on average $0.405 90% Switch a borrower choice of R to S 9 out of 10 times on average $0.500 100% Switch a borrower choice of R to S 10 out of 10 times Earnings: If they receive the loan, the earnings of the borrowers depend on the project choices made by the two borrowers and on the monitoring costs the two borrowers choose to incur. If the lender decides to make the loan, her earnings depend on the actual project choices made by the two borrowers. If she chooses not to invest in the loans to the borrowers, her money is allocated to a savings account and she earns $1.50 for the period. The earnings of the two borrowers and the lender in the different project scenarios are as follows. Here C1 and C2 denote the monitoring costs incurred by borrower 1 and 2 respectively. Actual project
choice of borrower 1
Actual project choice of
borrower 2
Earnings of borrower 1
Earnings of borrower 2
Earnings of lender
S S $1.75 – C1 $1.75 – C2 $2.50
40
S R $0.00 – C1 $2.50 – C2 $2.00 R S $2.50 – C1 $0.00 – C2 $2.00 R R $2.50 – C1 $2.50 – C2 -$2.00 No loan is provided $0.00 $0.00 $1.50
Each borrower can increase the chances of the other choosing project S by investing in monitoring. Monitoring choices will have to be made simultaneously and before each borrower knows whether the lender actually makes the loan. Examples: Consider the following examples, which were chosen randomly and are not meant to suggest any particular decisions. Example # 1: 1. Lender makes the loan. 2. Borrower 1 chooses project S and monitoring M of 70%. Monitoring cost C1 =
$0.245. 3. Borrower 2 chooses project R and monitoring M of 30%. Monitoring cost C2 =
$0.045 4. Monitoring results: Borrower 1’s monitoring is unsuccessful and so borrower 2’s
actual project is project R. Borrower 2’s monitoring is also unsuccessful, but borrower 1 already chose project S, and so his actual choice remains project S.
5. Earnings: Use the second row of the previous table to determine borrower 1’s earning = $(0.00 – 0.245) = –$ 0.245; borrower 2’s earning = $(2.50 – 0.045) = $2.445; and lender earning = $2.00
Example # 2: 1. Lender makes the loan. 2. Borrower 1 chooses project R and monitoring M of 80%. Monitoring cost C1 =
$0.320. 3. Borrower 2 chooses project S and monitoring M of 50%. Monitoring cost C2 =
$0.125. 4. Monitoring results: Borrower 1’s monitoring is successful, but borrower 2 already
chose project S and so his actual project choice remains project S. Borrower 2’s monitoring is also successful and this switches borrower 1’s actual project choice to S.
5. Earnings: Use the first row of the previous table to determine borrower 1’s earning = $(1.75 – 0.320) = $1.430; borrower 2’s earning = $(1.75 – 0.125) = $1.625; and lender earning = $2.50.
Summary of Decisions to be taken: Lender:
1. In every period choose how you want to invest your $2, using a decision screen shown in Figure 1.
41
Borrowers: 1. Indicate how much you wish to invest in monitoring the other borrower to
possibly switch him or her to project S, as shown in Figure 2, in case you receive the loans.
2. Decide whether you want to invest in project S or project R, using a decision screen shown in Figure 3. Remember that if the other borrower chooses to incur a monitoring cost (shown as other’s C on the figure), there is a chance of M that your project choice will be switched to S, even if you had actually chosen R.
Remember that choices are made simultaneously and the borrowers do not know whether the lender chose to invest in the loans or not before making their choices of a project. Once both the lender and the borrowers have made their decisions, the information shown in Figure 4 will be provided to all of the participants in the group: • Borrower’s project choices • Did the lender choose to make the loan • Borrower’s monitoring level, if the lender chose to make the loan • Actual projects chosen by the borrower, if the lender chose to make the loan • Lender earnings • Borrower earnings • Your cumulative earnings over the experiment Attached to these instructions is a record sheet where you are required to record your earnings and other details from every period. Are there any questions before we start the experiment?
42
Figure 1: Lender’s loan decision screen
43
Figure 2: Borrower’s monitoring decision screen
44
Figure 3: Borrower’s project decision screen
45
Figure 4: Example output screen
46
Quiz
Participant ID: ________ Total Number of Correct Answers: Earnings: Total Number of Correct Answers × $0.50 =
47
Record Sheet Participant ID: ________
Period Borrower 1’s Project Choice (circle one)
Borrower 2’s Project Choice (circle one)
Did you make Loan? (circle one)
Borrower 1’s Monitoring Level
Borrower 2’s Monitoring Level
Borrower 1’s Actual Project (circle one)
Borrower 2’s Actual Project
Your Earnings
Borrower 1’s Earnings
Borrower 2’s Earnings
Your Cumulative Earnings
1 S R S R Yes No S R NA S R NA
2 S R S R Yes No S R NA S R NA
3 S R S R Yes No S R NA S R NA
4 S R S R Yes No S R NA S R NA
5 S R S R Yes No S R NA S R NA
6 S R S R Yes No S R NA S R NA
7 S R S R Yes No S R NA S R NA
8 S R S R Yes No S R NA S R NA
9 S R S R Yes No S R NA S R NA
10 S R S R Yes No S R NA S R NA
11 S R S R Yes No S R NA S R NA
12 S R S R Yes No S R NA S R NA
13 S R S R Yes No S R NA S R NA
14 S R S R Yes No S R NA S R NA
48
Period Borrower
1’s Project Choice (circle one)
Borrower 2’s Project Choice (circle one)
Did you make Loan? (circle one)
Borrower 1’s Monitoring Level
Borrower 2’s Monitoring Level
Borrower 1’s Actual Project (circle one)
Borrower 2’s Actual Project
Your Earnings
Borrower 1’s Earnings
Borrower 2’s Earnings
Your Cumulative Earnings
15 S R S R Yes No S R NA S R NA
16 S R S R Yes No S R NA S R NA
17 S R S R Yes No S R NA S R NA
18 S R S R Yes No S R NA S R NA
19 S R S R Yes No S R NA S R NA
20 S R S R Yes No S R NA S R NA
21 S R S R Yes No S R NA S R NA
22 S R S R Yes No S R NA S R NA
23 S R S R Yes No S R NA S R NA
24 S R S R Yes No S R NA S R NA
25 S R S R Yes No S R NA S R NA
26 S R S R Yes No S R NA S R NA
27 S R S R Yes No S R NA S R NA
28 S R S R Yes No S R NA S R NA
29 S R S R Yes No S R NA S R NA
30 S R S R Yes No S R NA S R NA
49
Period Borrower
1’s Project Choice (circle one)
Borrower 2’s Project Choice (circle one)
Did you make Loan? (circle one)
Borrower 1’s Monitoring Level
Borrower 2’s Monitoring Level
Borrower 1’s Actual Project (circle one)
Borrower 2’s Actual Project
Your Earnings
Borrower 1’s Earnings
Borrower 2’s Earnings
Your Cumulative Earnings
31 S R S R Yes No S R NA S R NA
32 S R S R Yes No S R NA S R NA
33 S R S R Yes No S R NA S R NA
34 S R S R Yes No S R NA S R NA
35 S R S R Yes No S R NA S R NA
36 S R S R Yes No S R NA S R NA
37 S R S R Yes No S R NA S R NA
38 S R S R Yes No S R NA S R NA
39 S R S R Yes No S R NA S R NA
40 S R S R Yes No S R NA S R NA
Recommended